Add LLM long-term memory system

- Implement LLMMemory module with SQLite + FTS5
- Add conversation tracking with context awareness
- Track NPC reputation and interactions
- Integrate memory with LLMBrain for contextual replies
- Add comprehensive test suite (13 tests)
- Update settings with memory_db_path config

tests/test_llm.py

@@ -172,10 +172,11 @@ class TestLLMBrain:
     async def test_social_reply_with_mock(self):
         brain = LLMBrain()
         brain._available = True
-        brain.client.ask = AsyncMock(
+        # Mock generate since social_reply now uses it with context
+        brain.client.generate = AsyncMock(
             return_value=_make_mock_response("Greetings, fellow adventurer!")
         )
-        result = await brain.social_reply("Hello!")
+        result = await brain.social_reply("Hello!", sender="TestPlayer")
         assert result is not None
         assert "adventurer" in result.lower()